Integrating Non-Additive Genomic Relationship Matrices into the Study of Genetic Architecture of Complex Traits.

The study of genetic architecture of complex traits has been dramatically influenced by implementing of genome-wide analytical approaches during recent years. Of particular interest are genomic predictions strategies which mainly aim at making use of genomic information for predicting yet-to-be observed phenotypes instead of detecting trait-associated loci. In this work we present the results of a simulation study to improve our understanding of the statistical properties of estimation of genetic variance components of complex traits, and of additive, dominance and genetic effects through best linear unbiased prediction (BLUP) methodology. Simulated dense marker information was used to construct genomic additive and dominance matrices, and multiple alternative pedigree- and marker-based models were compared to determine if including dominance term into the analysis may improve the genetic analysis of complex traits. Our results showed that a model containing pedigree- or marker-based additive relationship matrix along with a pedigree-based dominance matrix provided the best partitioning of genetic variance into its components especially when some degree of true dominance effects was expected to exist. Also, we noted that the use of a marker-based additive relationship matrix along with a pedigree-based dominance matrix had the best performance in terms of accuracy of correlation between true and estimated additive, dominance and genetic effects.